I'm trying to define a new generic, and keep getting an S3 generic/method consistency when running R CHECK. All of the code seems to be working, and I'm not getting any note, errors, or other warnings.
This minimal example shows the warning I'm getting. The functions are
myscale <- function(x, y) UseMethod("myscale")
myscale.default <- function(x) x
The usage section of the man file is
\usage{
myscale(x, y)
\method{myscale}{default}(x)
}
and the NAMESPACE file is
export("myscale", "myscale.default")
S3method(myscale, default)
When I build the package and run CHECK, I get
* checking S3 generic/method consistency ... WARNING
myscale:
function(x, y)
myscale.default:
function(x)
See section 'Generic functions and methods' in the 'Writing R
Extensions' manual.
As I understand it, there shouldn't be a problem as long as the generic function contains all possible arguments of any method, and the methods have their arguments in the same order as the generic. It seems that having one method with only "x" shouldn't be a problem. I've read the section mentioned in the warning, but I can't figure out what's going on. Is any of this wrong?
Jay
[R-pkg-devel] R CHECK warning about new S3 generic/method consistency
7 messages · CRAN.r, Diego Hernangómez Herrero, Duncan Murdoch +1 more
Shouldn?t you include the y argument also in myscale.default ? Your generic is defining that argument as well. Have a nice day! El El lun, 11 mar 2024 a las 18:25, CRAN.r <cran.r at proton.me> escribi?:
I'm trying to define a new generic, and keep getting an S3 generic/method
consistency when running R CHECK. All of the code seems to be working, and
I'm not getting any note, errors, or other warnings.
This minimal example shows the warning I'm getting. The functions are
myscale <- function(x, y) UseMethod("myscale")
myscale.default <- function(x) x
The usage section of the man file is
\usage{
myscale(x, y)
\method{myscale}{default}(x)
}
and the NAMESPACE file is
export("myscale", "myscale.default")
S3method(myscale, default)
When I build the package and run CHECK, I get
* checking S3 generic/method consistency ... WARNING
myscale:
function(x, y)
myscale.default:
function(x)
See section 'Generic functions and methods' in the 'Writing R
Extensions' manual.
As I understand it, there shouldn't be a problem as long as the generic
function contains all possible arguments of any method, and the methods
have their arguments in the same order as the generic. It seems that having
one method with only "x" shouldn't be a problem. I've read the section
mentioned in the warning, but I can't figure out what's going on. Is any of
this wrong?
Jay
______________________________________________ R-package-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-package-devel
On Monday, March 11th, 2024 at 12:43 PM, Diego Hernang?mez Herrero <diego.hernangomezherrero at gmail.com> wrote:
Shouldn?t you include the y argument also inmyscale.default ? Your generic is defining that argument as well.
I assume (hopefully correctly) that methods don't need to include all the arguments of the generic. I get the same warning if I use "..." instead of y, too.
Usually when I write a generic I use ? in the definition to allow using
additional arguments in specific methods, see
as_bibentry <- function(x, ...) { UseMethod("as_bibentry")
}
as_bibentry.cff <- function(x, ..., what = c("preferred", "references",
"all")) {
<code>
}
but in any case the method must include at least the arguments defined in
the same order. Using ? in the generic allows me to include `what` arg with
no WARNINGs.
Have a nice day!
El El lun, 11 mar 2024 a las 19:13, CRAN.r <cran.r at proton.me> escribi?:
On Monday, March 11th, 2024 at 12:43 PM, Diego Hernang?mez Herrero < diego.hernangomezherrero at gmail.com> wrote:
Shouldn?t you include the y argument also inmyscale.default ? Your
generic is defining that argument as well. I assume (hopefully correctly) that methods don't need to include all the arguments of the generic. I get the same warning if I use "..." instead of y, too.
On 11/03/2024 2:13 p.m., CRAN.r wrote:
On Monday, March 11th, 2024 at 12:43 PM, Diego Hernang?mez Herrero <diego.hernangomezherrero at gmail.com> wrote:
Shouldn?t you include the y argument also inmyscale.default ? Your generic is defining that argument as well.
I assume (hopefully correctly) that methods don't need to include all the arguments of the generic. I get the same warning if I use "..." instead of y, too.
No, your assumption is backwards. The methods do need to include all arguments of the generic. As Writing R Extensions says near the start of section 7, "A method must have all the arguments of the generic, including ? if the generic does." Think about your user. They'll ask about help for `inmyscale`, and see that it has two arguments, x and y. If x is a type that goes to `inmyscale.default`, the user would receive an error when they followed the docs and included the y value. The usual way to handle this is to include both x and y in all methods, but document some of them to say that y is ignored. Duncan Murdoch
No, your assumption is backwards. The methods do need to include all arguments of the generic. As Writing R Extensions says near the start of section 7, "A method must have all the arguments of the generic, including ? if the generic does."
That's embarrassing. I was worried it was something simple I missed. Thanks for pointing that out!
On 11.03.2024 19:34, CRAN.r wrote:
No, your assumption is backwards. The methods do need to include all arguments of the generic. As Writing R Extensions says near the start of section 7, "A method must have all the arguments of the generic, including ? if the generic does."
That's embarrassing. I was worried it was something simple I missed. Thanks for pointing that out!
______________________________________________ R-package-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-package-devel
Even more embarassing given you seem to be Mr CRAN given your mail message's "From" field ... Uwe Ligges